Determination of atherosclerotic heart disease risk based on multiple clinical variables in men aged 40-65 years

Introduction: Atherosclerotic heart disease (ASHD) is a significant cause of death in Türkiye and worldwide. Early diagnosis of this disease, which can begin early in life and progress silently for a long time, is crucial for reducing mortality and morbidity. Currently available scoring systems, such as Systematic Coronary Risk Evaluation 2 (SCORE2) and Framingham, do not utilize sufficient parameters to predict the risk of ASHD. The aim of this study is to develop a model system that can better predict the risk of ASHD using new parameters found to be effective on ASHD and make it available in primary health care settings. Materials and Methods: This study was conducted with 350 men aged 40-65 who attended the Family Medicine outpatient clinic of Bilkent City Hospital in Ankara. Half of them (175) had atherosclerotic cardiovascular disease, and the other half served as controls. It is examined that factors such as demographic characteristics, lifestyle, blood pressure, and laboratory results. It is also calculated the Single Point Insulin Sensitivity Estimator (SPISE) and Atherogenic Index of Plasma (AIP) indices. ROC analyses and machine learning models were used for predictive modeling. Results: Older age, family history, obesity, high systolic blood pressure, high triglycerides, low HDL-Cholestrol (HDL-C) and high HbA1c were associated with atherosclerotic cardiovascular disease. SPISE values were lower in thE atherosclerotic cardiovascular disease group, while AIP values were higher. ROC analysis showed that AIP and SPISE better discriminated cases from those with normal lipid markers. The Support Vector Machine (SVM) model had the best accuracy (94%) and AUC (0.971) among machine learning algorithms. Conclusion: Newer variables such as SPISE, AIP, exercise level, pack-years, HbA1c and triglycerides are considered together with common risk factors, they significantly increase the accuracy of ASHD risk estimates. Incorporating these measurements into regular risk screening in primary care settings could facilitate early identification of at-risk individuals. Furthermore, the use of the Support Vector Machine model to determine ASHD risk has higher accuracy (AUC: 0.971) than standard methods. This suggests that data-driven methods capable of analyzing multiple parameters simultaneously could be valuable in clinical risk assessment when integrated into the HYP system used in family medicine.